Iguazio develops and provides a data science platform built to enable the acceleration and scaling of development, deployment, and management of AI applications with MLOps and automation of machine learning pipelines. The company serves a variety of industries, including financial services, telecommunications, smart mobility, manufacturing, retail, ad-tech, gaming, healthcare, and energy and utilities.
The platform enables the automation of machine learning pipelines, from data collection preparation and training to rapid deployment and ongoing monitoring in production. Unstructured and structured data can be ingested from any source in real time and unified. Subsequently, online and offline features can be built using Iguazio's Integrated Feature Store. Experimentation can be run over scalable serverless machine learning/data science runtimes with automated tracking, data versioning, and continuous integration/delivery support. Models and APIs can be deployed from a Jupyter notebook or an integrated development environment so production and model performance can be monitored.
Iguazio defines features as properties that are used as inputs to a machine learning model. Feature engineering involves the generation of new features used for training (based on historical data) and for model prediction (based on online or real-time data). The Iguazio integrated feature store, which is a central component of Iguazio's MLOps platform, aims to solve problems related to the process of feature engineering via automation.
According to Iguazio, its feature store is the first commercially available production-ready feature store that is part of an integrated and data science and engineering solution. The solution is designed to automate and simplify how features are engineered, with a single real-time and batch integration. In this process, high-level transformation logic is automatically converted to real-time serverless processing engines that can read from any online or offline source, handle any type of structures or unstructured data, and run computation graphs and native user code. Iguazio’s solution operates on a multi-model database, serving the computed features through a variety of APIs and formats, such as files, SQL queries, pandas, real-time REST APIs, time-series, and streaming.
The platform enables the creation of feature engineering processes with an integrated data transformation service, including feature aggregations with sliding windows, pre-built transformations, or custom logic in native Python code. Iguazio's goal is to provide a user-friendly API and SDK that data scientists can use to create features without requiring long data engineering cycles.
Features can be shared, searched, collaborated on, and evaluated with detailed statistics and analysis on the Iguazio platform, enabling users to gain insight into how features correlate to data sources and models.
In addition, statistics can be shown in real time, enabling drift detection based on actual data drift. The Iguazio feature store is fully integrated with other features of the Data Science Platform, such as concept drift monitoring and feature monitoring.
Features are developed once for both offline and real-time deployment. Based on the information contained in incoming events or streams, the feature transformation pipeline processes features in real time and serves the results at millisecond level latency or forwards them directly into a stream.
The platform saves the data history of features, with the tracking information capturing how each feature was generated for regulatory compliance purposes. Completed models can be deployed in a hybrid multi-cloud environment.